BetaWeekly digests are a beta — we're trying something new. Feedback welcome.
Figma's agent goes GA as Linear tightens AI spend controls
October 5–11, 2026
The Figma agent left beta with faster multiplayer performance and design-library-aware tooling, while Linear made it easier to unblock stacked pull requests and put hard limits on AI credit spend.
The week at a glance
Replay the weekBiggest releases
- 01Figma agent exits beta with lower latency and design library tools
Figma - 02Linear adds stacked pull request support in Diffs and AI credit spend controls
Linear - 03Notion 1.7.345 bug fixes and performance improvements
Notion - 04Linear Mobile 1.103.0 bug fixes and minor improvements
Linear - 05Linear Mobile 1.102.0 bug fixes and minor improvements
Linear
Design agents grow up
Figma's biggest move of the week was taking its agent out of beta. The headline is latency — responses now arrive fast enough to feel conversational rather than queued — but the more consequential change is multiplayer. Multiple people can now direct the agent in the same file without trampling each other, which finally makes it usable in the shared, messy files where design work actually happens. Figma also widened the agent's access to file context and shipped new tools that guide how it reads and applies design libraries, so generated work lands closer to the team's existing components instead of drifting into one-off styles. For teams that had been treating the agent as a novelty, the combination of speed and library awareness is the point at which it starts fitting into a real workflow.
Stacked PRs and AI spend controls in Linear
Linear's week split between code review and cost governance. On the review side, stacked pull requests now render in Linear Diffs: walking a stack shows review, CI, and merge status at a glance, related PRs collapse into a single row alongside their issue and project context, and GitHub-native stacks can preview every affected PR before you merge the current one and everything beneath it. Diffs also update in real time as the agent works, so you watch changes land during a coding session rather than refreshing for them.
The other half is money. Admins get an AI credit usage explorer that breaks spend down by team, feature, user, or loop, plus spend limits at the workspace, user, or loop level that pause additional usage once a ceiling is hit — a sensible answer to the question every org adopting agents eventually asks. Guided Reviews, previously gated higher, also reached Basic plans. Linear Mobile picked up two rounds of minor fixes, nothing that changes how the app behaves day to day.
Housekeeping elsewhere
Notion pushed bug fixes and performance improvements, and Discord's patch notes collected the usual sweep of performance, reliability, and usability fixes. Slack had a rougher week: the EMM build got stuck in a sign-in loop when the main Slack app was also installed, now fixed, followed by a further round of fixes to close out the week. Nothing here will change how you work, but the EMM fix matters if your organization deploys Slack through managed distribution.